Experiment Tracking (MLflow)¶

Purpose¶

MLflow is used to track: - experiments, - metrics, - parameters, - artifacts.

It provides a single pane of glass for ML iteration.

Logged artifacts¶

Each run logs: - model parameters, - validation metrics, - feature importance or diagnostics, - config snapshots, - evaluation reports.

Traceability¶

Each MLflow run is traceable to: - git commit hash, - DVC dataset version, - training configuration.

This enables full experiment auditability.